v1.2.0

Added

• Output signature standard (docs/output-standard.md) defining required artifact sectioning and a mandatory Decision Frame block.
• Deterministic recovery playbooks (docs/recovery-playbooks.md) mapping failure triggers to exact remediation steps.
• Binary quality gates (evals/pass-fail.md) for pass/fail enforcement before rubric scoring.
• Direct "Why this beats raw AI" comparison framing in README.

Changed

• Core skill enforcement: PRD, Strategy, Design Review, and A/B Analysis skills now require the Shipwright signature structure.
• Validation hardening: scripts/validate.sh now checks pass/fail gate presence, output/recovery docs, and signature section presence in core skills.
• Evaluation flow: evals/README.md now enforces pass/fail-first, scoring-second.
• Composition model clarity: explicit primitives (Workflow = graph(Skills), Agent = constrained executor, Orchestrator = planner over graph).
• Cross-tool guidance: pass/fail gating and signature usage added for standalone mode.